Claude vs ChatGPT for Google ads - which is best?

Claude vs. ChatGPT: Which is Better for Marketing and Google Ads?

Introduction to the Comparison

• The video compares Claude Code (Opus 4.6) and Codex (GPT 5.4) in terms of their performance on marketing tasks, specifically Google Ads audits and ad copy generation.

Experiment Setup

• Both models were given the same prompt to conduct a search term waste audit, build a client audit report, and write fresh ad copy for a Google Ads account.

• The environment was controlled using VS Code to ensure fair testing conditions with identical instructions provided to both models.

Performance Observations

• Claude Code completed the task significantly faster than Codex, finishing in 5 minutes compared to Codex's longer duration due to some sub-agent failures during execution.

• Despite different speeds, both models produced varying insights and analyses from the same prompts, highlighting their distinct approaches to problem-solving.

Results of the Audit Reports

Report Presentation

• Both models generated HTML reports; however, Codex's report was styled more closely aligned with the website's design aesthetics despite not being prompted for it explicitly.

• Claude’s report had interactive elements but lacked visual appeal compared to Codex’s output which surprised the presenter given previous assumptions about Claude's design capabilities.

Depth of Analysis

• Claude provided a more detailed analysis in its audit report with better readability and depth regarding campaign performance metrics compared to Codex’s top-level stats that were less informative.

• Key findings from both reports included similar insights about mobile conversions but differed in detail; Claude identified deeper issues such as conversion tracking needs and impression share opportunities while Codex remained surface-level in its analysis.

Search Term Waste Audit Insights

Comparative Findings

• In assessing wasted spend, ChatGPT offered general insights that were easier for clients or stakeholders unfamiliar with Google Ads terminology but lacked specific actionable recommendations compared to Claude’s detailed breakdown of waste by intent theme and ad group performance metrics.

• While both models identified relevant terms that weren't converting effectively, they also made mistakes regarding certain keywords' relevance which emphasized the importance of human oversight when interpreting AI-generated results.

Ad Copy Generation Evaluation

Quality of Ad Copy

• In generating RSA ad copy, Claude outperformed ChatGPT significantly by providing compelling headlines that addressed urgency and trust factors effectively while ChatGPT's suggestions were deemed bland and lacking specificity or creativity.

• Claude demonstrated an understanding of what converts best by analyzing data before crafting headlines whereas ChatGPT failed to incorporate critical feedback into its outputs leading to generic suggestions without urgency or clarity on benefits offered by services advertised.

Conclusion on Model Performance

Overall Assessment

• The final scores reflected a clear advantage for Claude over Codex: 8.7 versus 7 respectively based on task completion quality where Claude delivered comprehensive feedback without skipping requested details unlike Codex which missed nuances despite running longer processes during execution timeframes.

Future Considerations

• Emphasis was placed on developing skills within AI systems for improved accuracy in outputs; viewers are encouraged to utilize shared resources for enhancing their own interactions with these tools moving forward as new features are anticipated soon including autonomous agents tailored specifically for paid marketers across various platforms beyond just Google Ads.